Ayush Agarwal

Results-driven software engineer with a strong background in data science, generative AI, and test automation. Proven track record of leading successful client POC, from user story to manual test case and automation script generation, utilizing NLP and data processing techniques. Skilled in developing Python-based automation frameworks using Selenium, including an innovative Auto Healing tool that significantly reduced troubleshooting time. Experienced in researching and implementing cutting-edge technologies like Qwen2.5-VL and Claude models for FSD-to-user-story automation, showcasing a commitment to driving innovative, data-driven impact through ethical AI solutions. Passionate about exploring Agentic AI and seeking opportunities as a data scientist or AI engineer to continue making a meaningful contribution in the field.

Auto Healing Framework (Python, Selenium, Gradle, Neo4j)

 - Built a tool to dynamically monitor exceptions and implement fixes, using Gradle for build automation.

 - Impact: Reduced manual troubleshooting time by 80%.

Agentic AI Vision-Enhanced Process Automation (Python, Selenium, GPT-4o)

 - Developing a Selenium-based crawler with GPT-4o vision to automate UI-driven processes, exploring MultiOn-inspired browser automation.

 - Impact: Enhancing process automation robustness.

Manual Test Case to Automation Script with Shadow DOM Support (Python, Java, Selenium, JavaScript, ChromaDB)

 - Automated manual test case conversion to Selenium Java scripts using LLMs, with a crawler supporting Shadow DOM via JSPath.

 - Processed client test cases to generate action objects, maintaining XPath/JSPath mapping files. - Impact: Streamlined test automation by 50%.

End-to-End Generic Manual Test Case Automation (Python, JavaScript, ChromaDB, LLMs)

 - Built a generic crawler to process manual test cases, extract elements, and store in ChromaDB, using cosine similarity for action steps.

 - Impact: Improved automation scalability.

Data Extraction with LangChain and Azure (Python, LangChain, Azure)

 - Processed unstructured client documents with LangChain and Azure using NLP to generate structured outputs for decision-making.

 - Impact: Enhanced data processing efficiency.

User Story to Manual Test Case PoC (Python, ChatGPT, ChromaDB)

 - Processed historical client data with ChatGPT and ChromaDB cosine similarity to generate manual test cases from user stories.

 - Impact: Streamlined test case creation.

Vision-Language Model Fine-Tuning for UI Automation (Python, Unsloth, PyTorch)

 - Researched and fine-tuned Qwen2.5-VL (2B) on UI screenshots and test steps to automate test case generation for login and data table scenarios.

 - Impact: Reduced test case creation time by 45%.